[WikiEN-l] Meaningful estimates of popularity?

[geni]

On 24/11/2007, David Gerard <dgerard at gmail.com> wrote:
> The Wikicharts are back in the news, with BoingBoing reporting on
> Conservapedia's most popular pages:
>
> http://www.boingboing.net/2007/11/21/top-ten-most-viewed.html
>
> The Wikicharts are a sampled raw page count:
>
> http://hemlock.knams.wikimedia.org/%7Eleon/stats/wikicharts/index.php?lang=en&wiki=enwiki&ns=alle&limit=100&month=11%2F2007&mode=view
> http://meta.wikimedia.org/wiki/WikiCharts
>
> The hits on "Main Page" will be people just going to the site. The
> hits on "wiki" will not be because of a society-wide fascination with
> editable websites, but (I would guess) because people have typed
> "wiki" into a search engine. I suspect a lot of the sex-related hits
> will be disappointed porn seekers.

Doubtful. Consider the most looked up words in your average kid's
dictionary. In addition someone looking for info on armored fighting
vehicles could be looking at one of hundreds of pages (anything from
[[Bob Semple tank]] to [[Iosif Stalin tank]]) people looking for stuff
on sex are likely to end up at one of of a smaller number of pages
thus the interest ends up being concentrated over a small number of
pages.

> The question is: what does "popular" actually usefully mean? Raw page
> hits demonstrably isn't quite it. "Pages with most hits gone to by
> people looking for information" (whether starting from Wikipedia, a
> search engine or a link) is closer to what we're after (even if it's
> hard to quantify intent). Ideas?

The normal approach is to look at viewing times (for example we know
people tend to look at Barack Obama's article for longer than they
tend to look at Hillary Clinton's). However we can't do that. The next
approach would be to look at how many related articles people look at
but again there would be privacy issues so we can't really do that.

-- 
geni